4
$\begingroup$

I am wondering whether some criterion can be put on a category $\mathcal{C}$ with direct sums to ensure that for three objects $X,Y,Z$ one has $$ X\oplus Y \cong X \oplus Z\Longrightarrow Y\cong Z. $$ For $\mathcal{C}$ being the category of finitely generated abelian groups this is certainly true, and this is also true for arbitrary abelian groups when $X$ is finitely generated by a theorem of Cohn and Walker. In the case that $\mathcal{C}$ is the category of $\mathbb{Z}[G]$-modules, where $G$ is a finite group, this property is discussed at length in Richard Swan's paper Projective modules over binary polyhedral groups J. Reine Angew. Math. 342 (1983), 66–172, MR0703486.

I would expect some formal criterion at least to exist, even if it may be relatively hard to check or to apply. Of course, the more useful and easier to apply, the happier I am!

$\endgroup$
6
  • 2
    $\begingroup$ For finitely-generated abelian groups, or finitely-generated modules over a PID, doesn't this follow directly from the classification thereof? Perhaps the theorem of Cohn and Walker is meant to refer to something else? $\endgroup$ Commented Apr 30, 2021 at 22:28
  • $\begingroup$ @TimCampion Thanks, you are right! I have modified and corrected the statement. $\endgroup$ Commented May 1, 2021 at 18:49
  • $\begingroup$ I edited in a link to the Swan paper. Do you have a reference for the theorem of Cohn and Walker? $\endgroup$
    – LSpice
    Commented May 1, 2021 at 18:51
  • $\begingroup$ I'm more familiar with formal non-cancellation arguments like the Eilenberg swindle. Even when everything is "suitably finite", non-cancellation shows up when talking about stably-free modules and bundles, for example. OTOH I suppose if everything is free and finitely-generated, then one is talking about Invariant Basis Number rings, among which are all commutative rings. I think cancellation may also hold in an Artinian abelian category? $\endgroup$ Commented May 1, 2021 at 18:51
  • 2
    $\begingroup$ You might need to look at Krull-Schmidt categories, they satisfy this property. $\endgroup$
    – user127776
    Commented May 1, 2021 at 18:54

0

You must log in to answer this question.

Browse other questions tagged .